Welcome to TAS 2026: Trustworthy Agentic Systems, a symposium in the AAAI 2026 Fall Symposium Series. Agentic systems, in which large language models (LLMs) plan, invoke tools, retrieve external knowledge, and act autonomously over multiple steps, are moving from research prototypes into consequential, real-world decision making. Our ability to guarantee that these systems are reliable, private, safe, and accountable has not kept pace. TAS convenes the artificial intelligence community to treat responsibility as a first-class design objective that is engineered, measured, and governed across the full lifecycle of an agent. The symposium is organized around two complementary themes:
TAS 2026 • Part of the AAAI Fall Symposium Series
Dates: November 5–7, 2026 Location: Westin Arlington, Arlington, Virginia, USA Format: In person (per AAAI policy) Contacts: sumon@case.edu, adas@uakron.edu, shahnewazkarim-sakib@utc.edu, shibbir@txstate.edu
Note: This site is under construction. Items marked TBA will be updated as the program is finalized.




Full bios are available on the Organization page.
TAS welcomes full papers, short papers, position statements, and one-page abstracts, as well as statements of interest from prospective attendees. Speculative work and work in progress are encouraged alongside completed results.
Submissions are formatted using the AAAI-26 author kit.
The “Preferred” submission round, with the option to be included in the AAAI proceedings:
Regardless of archival plans, authors are encouraged to submit earlier rather than later. All participants (presenters, organizers, and attendees) are required to register for the event. Full details are available on the Call for Participation page.
TAS will feature invited speakers, paper and spotlight sessions, target-problem working groups, an open-format panel, and a tool and dataset showcase, spread across a two and one-half day program. See the full schedule on the Schedule page.